Did a Machine Just Crack Math's Most Famous Puzzle? A Human Says He Did Too
Who gets the credit when one of the hardest problems in mathematics falls twice in the same week? That's the strange question hanging over the math world this Friday.
Levent Alpöge, a mathematician who works at Anthropic, says he's found a solution to the Riemann Hypothesis, the 1859 puzzle about the hidden pattern behind prime numbers. It's one of the Millennium Prize problems, the kind of thing people spend whole careers circling. His version was built with help from an AI model and written so it compiles in Lean, a language that machine-checks every logical step. In an essay about it, he admitted the timing was "horrible" but said he couldn't bring himself to sit on it. He said the model "flew, and danced, and sparkled" as it closed in on the finish line. He and NYU's Tristan Buckmaster also released a proof tied to the Navier-Stokes equations.
Then came the twist. Stanford professor Ravi Vakil posted a draft proof of the General Riemann Hypothesis by Stanford postdoc Alessandro Masullo, stressing that it's "all human, with no AI." Masullo says it took him twelve years. His message was basically, "We're still here, and we're not going anywhere."
Neither claim is verified yet. TechCrunch says most people following along seem fairly confident the Lean version will hold up, since the software has already checked the logic, but it still lacks a human-readable explanation, which Alpöge has promised to write. Masullo's proof is an early draft that was meant to be polished before going public.
If both check out, it's a wild moment, and it feeds a bigger contest that's been heating up since the summer, with AI labs and mathematicians racing to knock down famous open problems.
So does it matter who, or what, solved it first, as long as it's right?
An AI-assisted proof and a 12-year human effort both claim the Riemann Hypothesis. If both hold up, who deserves the credit?